Scenario-Aware Text Classification With Semantic Noise Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Deep learning-based text classification methods are affected by data quantity and training iterations, leading to accuracy issues and noise introduction.

Innovation Solution

A text classification method that incorporates semantic enhancement processing and non-linear mapping based on scenario information to improve text encoding, using initialized models and updating parameters for enhanced accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If deep learning models are used for text classification, then classification capability is improved, but accuracy is affected by data quantity and training iterations

Engineering Contradiction:

Improveclassification capabilityVSAvoidclassification accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent applies semantic enhancement processing to the input text before encoding it, which is a preliminary action that enriches the text representation. This semantic enhancement uses scenario information to augment the text with relevant contextual knowledge, improving the quality of the input before it reaches the classification model, thereby reducing dependence on large data quantities and extensive training iterations

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent transforms the input text into an enhanced semantic representation by changing its parameters through semantic enhancement processing. This involves enriching the text with additional semantic features and scenario information, effectively changing the parameter space of the input data to improve classification accuracy without requiring proportional increases in data quantity or training complexity

Inventive Principle:

Principle #35Parameter changes

2Speed

If text classification is performed directly on text encoding, then processing speed is maintained, but noise in the input text affects accuracy

Engineering Contradiction:

Improveprocessing speedVSAvoidclassification accuracy

Core Design Contradiction:

SpeedVSMeasurement precision

Solution Approach 1:

The patent introduces semantic enhancement processing as an intermediary step between the raw text input and the encoding process. This intermediary component filters and enriches the text representation, removing noise and adding relevant semantic information before the text is encoded and classified, thereby improving accuracy without significantly compromising processing speed

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If semantic enhancement processing is applied, then text representation is improved, but processing complexity increases

Engineering Contradiction:

Improvetext representation qualityVSAvoidprocessing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The semantic enhancement processing is applied locally to the specific portions of text that benefit most from enhancement, rather than uniformly processing the entire text. This selective approach improves text representation quality where needed while minimizing the overall processing complexity and computational overhead

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20250378277A1Text classification

Publication Date: 2025.12.11 MASHANG CONSUMER FINANCE CO LTD

AI summary

In a text classification method, an input text and scenario information corresponding to the input text is received. Semantic enhancement processing on the input text is performed based on the scenario information to obtain a semantic enhancement result. The semantic enhancement result is encoded to obtain text encoding. Non-linear mapping processing is applied on the text encoding to obtain text classification results of the input text.